lpem-feb ui working paper 018, march 2018 issn 2356-4008 · are served by pdam of tangerang...

LPEM-FEB UI Working Paper 018, March 2018ISSN 2356-4008

Estimating Urban Water Demand Elasticities usingRegression Discontinuity: A Case of TangerangRegency, IndonesiaMuhammad Halley Yudhistira1F, Prani Sastiono1, & Melly Meliyawati1∗

AbstractWe estimate the effect of water tariff adjustment in Tangerang city, Indonesia in November 2014 on monthly waterconsumption. Due to typical water-block pricing strategy, estimating water demand elasticities are likely to be complex.A unique panel monthly water consumption dataset at consumer level in Tangerang regency covering the period ofJanuary 2011–September 2016 is used. Using regression discontinuity framework, we find a 13% average tariff increasereduces 4% household water consumption on average. Further, our estimates suggest the tariff adjustment provides noeffects on high-income households, industrial, and commercial consumers. We also find more elastic response of waterconsumption in short-run period than in long-run.

JEL Classification: L95; R22; R53

KeywordsWater Demand Elasticities — Urban Water — Regression Discontinuity Design — Indonesia

1Institute for Economic and Social Research, Faculty of Economics and Business, Universitas Indonesia (LPEM FEB UI)FCorresponding author: Institute for Economic and Social Research (LPEM) Universitas Indonesia. Campus UI Salemba, SalembaRaya St., No. 4, Jakarta, 10430, Indonesia. Email: [email protected].

1. Introduction

Clean water plays important roles in fulfilling human basicneeds and the access is one of parameters for the qualityof life of society. Some manufacturing activities, like foodand beverage producers, also consider water as one of im-portant production inputs. Despite its importance, assessto clean water provision is lacking in Indonesia. By 2015,about 14% of Indonesian had unimproved drinking watersource (Patunru, 2015). The level of service is also lowsince regional water companies (Perusahaan Daerah AirMinum/PDAM), which provide piped water service covera small portion of water provision service (Asian Devel-opment Bank/ADB, 2016). The majority of water sourcecomes from unpriced groundwater that is harmful towardenvironment.

Growing population along with low level of piped wa-ter service and excessive groundwater use will worsen thesupply-demand gap of fresh water provision at nationallevel. Challenges in fresh water scarcity are likely to begreater in fast-urbanized areas because of fierce competi-tion among economic activities (Lundqvist et al., 2003).Understanding the demand behavior is needed to controlthe demand, while at the same investment in piped waterfacilities is also indispensable. Grasping how sensitive thedemand for water on its price will enable us to use the priceas potential instruments to manage the increasing demandover the time.

In this paper, we estimate the water demand elasticity us-∗We thank PT. Aetra Air Tangerang for data and discussion, and Ham-

dan Bintara and Wahyu Pramono for research assistance. We also thankAndhika Padmawan for his comments and inputs. Usual terms apply. Thisresearch received funding from the Institute for Economic and SocialResearch, Universitas Indonesia.

ing monthly water consumption of PT Air Aetra Tangerangconsumers in Tangerang Regency, Indonesia. The companyis privately owned, and together with the regional watercompany of Tangerang provide fresh-water service for sub-stantial number of citizens of Tangerang. Our monthly waterconsumption covers 57 months and more than 50 thousandcustomers forming about 1.6 million monthly-consumerpanel observations. The rich dataset also covers severaltypes of consumers, including household, business, com-mercial, and social, which enables us to estimate the con-sumption effect of tariff change across group of consumers.

To obtain the causal impact, we exploit event of un-precedented tariff adjustment in November 2014 as the forc-ing variable of regression discontinuity framework. Thecompany announced the tariff adjustment without briefannouncement in advance. We adopt Luechinger & Roth’(2016) estimation strategy to reduce the estimation bias bycontrolling consumer fixed-effect, polynomial time trend,and other control variables. Estimated standard errors areclustered to reduce potential bias due to serial correlation(Bertrand et al., 2004; Cameron et al., 2011). In the contextof Indonesia, we extend the work of Rietveld et al. (2000)who estimated residential water demand in Solo, Indonesiato obtain non-residential water demand elasticity as well asdistinguish between short-run and long-run estimates.

This study is also of international interest in that, typ-ical water demand elasticity estimations are potentiallybiased because of complex structure of water service tar-iff schemes and whether consumers fully aware with theschemes (Nataraj & Hanemann, 2011). We use regressiondiscontinuity with time as forcing variable to minimize theestimation biases, exploiting the fact that the percentagechange in tariff is relatively homogenous across blocks and

1

[email protected]

Estimating Urban Water Demand Elasticities using Regression Discontinuity: A Case of Tangerang Regency,Indonesia — 2/10

Figure 1. Map of Tangerang Regency and Sub-districts served by PT Aetra Air TangerangNote: grey areas indicate subdistricts served by PT Aetra Air Tangerang

Source: Direct correspondence with PT Aetra Air Tangerang (Geospatial Information Board, 2017)

consumer groups.Our paper proceeds as follows. Section 2 explains cur-

rent system of water provision in the Tangerang Regencyand how PT Aetra contributes in the system. Section 3describes our model and data, and discusses the expectedeffects of sudden tariff adjustment on water consumption.Section 4 provides discussion on the estimation results. Fi-nal section concludes.

2. Context: Water Utility Provision andPosition of PT Aetra Air Tangerang in

Tangerang Regency

Tangerang regency is situated in the Java island, the provinceof Banten, and 40 kilometers away to the west side ofJakarta, the capital city of Indonesia. The regency is partof greater metropolitan area, known as Jabodetabeka oneof the world’s largest urbanized areas. By 2016, there wereabout 823 thousand households living in Tangerang regency,spreading in 29 sub-districts. Most households are living inPasar Kemis sub-district, while the least households are inMekarbaru sub-district.

As commonly practiced in Indonesia, water provisionfor household and non-household activities in Tangerangregency mainly come from piped water provided by a re-gional water company (PDAM Tirta Kerta Raharja), or fromground water, river, and other natural source of water. Insome areas, private companies enter the market to providethe service. To extent the coverage of piped water provision,the local government of Tangerang regency invited privatecompany by 2008, i.e. PT Aetra Air Tangerang, to serve thedrinking water service in eight sub-districts over 25 years ofconcession. The eight sub-districts include Sepatan, Sepa-tan Timur, Pasar Kemis, Cikupa, Sindang Jaya, Sukamulya,Balaraja, and Jayanti using the water source from Cisadaneriver (see Figure 1). The company is designed to cover about72.000 households, equivalent of 350.000 inhabitants, andnon-household connections. Meanwhile, other sub-districts

are served by PDAM of Tangerang regency.Table 1 provides a recent tariff scheme as of March

2017. PT Aetra Air Tangerang imposes block pricing tariff;the higher water consumed, the higher per m3 water tariffto be paid. The tariff is also differentiated across consumers.Households, as the main consumer, are divided into fourgroups based on assets that primarily surveyed before theysubscribed into the service. Households with least asset willfall into R1 group and be charged with subsidized tariff. Thehighest is borne by R4 group which owns most assets. Non-household costumers are differentiated in similar fashion.Social consumers, such as mosques and orphanages aresubsidized. Other groups include government, commercial,and industrial use are charged much higher than the socialgroup.

The tariff system is, however, strictly regulated by thegovernment. At national level, Minister of Home AffairsRegulation No. 71 year 2016 on the Calculation and Deter-mination of Drinking Water Tariffs stipulates the maximumlevel of water tariff such that the cost borne by household forfirst 10 m3 consumption should not exceed 4% of Provincialminimum wage income. The company also cannot increasethe tariff except there is approval from the mayor. Anychange in tariff must be at least 2 years after the latest tar-iff adjustment. By far, the company has adjusted the tarifftwice, by October 2014 and March 2017. The first tariff ad-justment was more sudden and less anticipated by customersas the company announced the change after the new tariffwas imposed. The last adjustment was potentially moreanticipated as the plan had been socialized beforehand.

3. Method

Basic Model and Data We are interested in estimating theimpacts of the October 2014 water tariff adjustment onmonthly water consumption. To obtain the causal relation-ship, we use regression discontinuity with time variable asassignment variable (Auffhammer & Kellogg, 2011; Chen& Whalley, 2012; Luechinger & Roth, 2016). This approach

LPEM-FEB UI Working Paper 018, March 2018


Table 1. Water Tariff Regime by Consumer Type by March 2017

Consumer type Tariff scheme (Rupiah/m3)<10 m3 10–20 m3 >20 m3

Social 1450 1450 1450Government 10280 10280 10280Household

R1 2170 2170 2170R2 4530 5350 6440R3 7420 8905 10415R4 10400 12010 13450

Commercial 13300 14700 16100Industry 16726 16726 16726

Source: Keputusan Bupati Tangerang 690/Kep.206.Huk/2017

identifies the effect of the policy by examining the discontin-uous variation around the known cutoff (Jales & Yu, 2017),in our case is the month of new tariff implemented. Theestimation can be written as:

(1)ln(yit + 1) = αi + βDt + X′γ + θ f (t) + eit

The dependent variable is monthly water consumptionof consumer i in month t, presented in natural logarithm.The consumption is added by one to avoid dropped observa-tion under logarithm transformation. Water consumption isfrom metered monthly water consumption data of PT AetraAir Tangerang customers recorded from January 2012 toSeptember 2016, or for 57 months. With a total customerbase of 55.7 thousand subscribers, there are more than 3.2million potential observations. However, as some customershave not subscribed since January 2012, it may cause miss-ing observations, thus about 1.7 million observations can beused in the model.

Our variable of interest is dummy variable Dt , that cap-ture policy change in tariff of water. The value is 1 for theperiod after the new tariff is applied, and 0, otherwise. Weassume that the effect of change in price has started fromNovember 2014 as PT. Aetra Air Tangerang announcedthe policy by mid-October 2014. Percent change in waterconsumption due to change in the water price is then calcu-lated as 100(eβ −1) (Wooldridge, 2010). The elasticity ofeach consumer group k, εk, is then calculated by dividingthe percent change in water consumption by the percentchange in water price for respective group, θ k, such that

εk = 100(eβ−1)θ k .

Table 2 presents the water tariff regime by consumertype, before and after October 2014’s tariff adjustment.Tariff is discriminated by consumer groups and consump-tion size under cross-subsidy spirit. Consumers are dividedinto household and non-household consumers. Householdconsumers are further divided into four groups, based ontheir assets and/or land size. Per m3 tariff is higher forhigher household type, where R1 pays subsidized tariff.Non-consumer groups consist of subsidized social, govern-ment, commercial, and industrial consumers.

By October 2014, PT Aetra applied sudden adjustmentin tariff. Consumers, on average pay 13–15% higher tariff,except social and R1 household consumers. The companycan adjust the tariff at least biannually and must be consultedby the major. The adjustment must follow the principle ofaffordability and fairness. According to Minister of HomeAffairs Regulation No. 71/2016 on the Drinking Water Tar-iffs Calculation, household expenditure to fulfil drinking

water standards of basic needs (10 m3) should not exceed4% of Banten’s Provincial Minimum Wage. For example,the minimum wage of Banten Province is Rp3.021,650 by2016, as stipulated in the Decree of the Governor of Banten,No. 561/Kep-Huk/2016. concerning the determination ofthe minimum wage of Banten Province, the drinking watertariff should not exceed more than Rp71.360 in the event ofa tariff increase.

Our strategy to estimate the price elasticity relies onassumption that discontinuous price scheme changed atOctober 2014 is the only factor captured by variable Dt . Toensure this assumption, Equation (1) contains observablecontrol variables in X′ and relies on polynomial order toreduce the bias from unobservable factors.

X′ in Equation (1) is a vector containing list of controlvariables. We control for monthly GDP, which serves asincome variable. The variable is proxied by quarterly GDPdata and industrial production index. Our set of variablesalso include monthly precipitation at village level to controlactivities which need water use, such as garden watering.Impact from tariff change is also controlled through monthlyconsumer price index. We also introduce dummy variablesfor Idul Fithri celebration, which is held following lunarcalendar. We include dummy for drought in Cisadane river,which is served as source of Aetra’s water, in August 2015.Our dummies also include month-of-year and year dummiesto capture seasonality and different long-term trend, respec-tively. Detail explanation and source of data are presentedin Appendix A.

Due to data limitation at individual level, the model can-not control unobservable individual factors that potentiallyaffect water consumption, such as number of householdmembers, that may change during our data period. We useconsumer fixed-effect panel estimation to eliminate omit-ted variable bias from unobserved time-invariant variables.Additionally, many other unobserved variables, which varyover time, are not included into the model. Therefore, weintroduce polynomial time trend, f (t), to reduce the bias(Luechinger & Roth, 2016). We later perform a robustnesstest of polynomial order to check in what extent our elastic-ity estimates remain robust for the higher order.

Water consumption is unlikely to be independent acrosstime. In addition, our dataset involves relatively long series,i.e. 57 months. Both factors potentially create serial cor-relation problem (Bertrand et al., 2004). Hence, we allowunobserved disturbance eit to be correlated within sameconsumer and year, following Cameron et al.’s (2011) pro-cedure to capture within-consumer correlation and serial



Table 2. Water Tariff Regime by Consumer Type

Consumer typeTariff scheme (Rupiah/m3)

Before October 2014 After October 2014, before March 2017 Percent change0–10 m3 10–20 m3 >20 m3 0–10 m3 10–20 m3 >20 m3 0–10 m3 10–20 m3 >20 m3

Social 1.250 1.550 1.825 1.325 1.650 1.950 6.0 6.5 6.9Government 8.100 8.100 8.100 9.175 9.175 9.175 13.3 13.3 13.3Household

R1 1.900 1.900 1.900 2.025 2.025 2.025 6.6 6.6 6.6R2 3.575 4.150 5.000 4.045 4.775 5.750 13.2 15.1 15.0R3 5.850 6.950 8.100 6.625 7.950 9.300 13.2 14.4 14.8R4 8.125 9.250 10.375 9.200 10.625 11.900 13.2 14.9 14.7

Commercial 10.325 11.450 12.575 11.675 12.900 14.200 13.1 12.7 12.9Industry 13.216 13.216 13.216 14.934 14.934 14.934 13.0 13.0 13.0

Source: PT Aetra Air Tangerang (http://aat.co.id/informasi-rate/ retrieved at 2016)

correlation problem.The reported standard errors are thenclustered on each consumer-year combination.

The estimates will also suffer parameter bias if the con-sumers can anticipate the policy by adjusting their behaviorbefore the new tariff announced (see Lee & Lemieux, 2010).The response parameter is potentially underestimated if con-sumers may alter their water consumption behavior beforethe tariff adjustment was announced. In our setting, the biascan be, however, reduced as PT Aetra Air Tangerang an-nounced the new tariff once it was effectively implemented1.

Our dataset captures all PT Aetra Air Tangerang’s con-sumers, not only households, but non-households as well.This enables us to estimate Equation (1) for each consumergroups and compares the response across groups.

3.1 Short-run ResponseWe are also interested in examining whether the responsediffers across time frame. Demand can be more elastic orinelastic, depending on the nature of goods, in the short-run.To obtain the response of the first n-month, we modify anasymmetric model, which divides independent variable ofinterest into two states (York, 2012; Burke et al., 2015). Inshort, we estimate:

ln(yit +1) = αi+β1D1t +β2D2

t +X′γ +θ f (t)+eit (2)

where D1t is dummy variable taking values equals one for

first several months after tariff increase and zero, otherwise.Meanwhile dummy D2

t is assigned as zero for before tariffincrease and first several months in D1

t , and one, otherwise.This setting allows us to separate the average effect for thefirst several months (β1) and the effect thereafter (β2). Weestimate Equation (2) for first two until twelve months afterthe tariff increase.

3.2 Expected Response of the Water Tariff Adjust-ment

What is the impact of water tariff increase on monthly waterconsumption of PT Aetra Air Tangerang customers? Thebasic economic theory of demand states that tariff increaseinduces the water consumption reduction. The percentagereduction, nevertheless, is expected to be relatively smallerthan the percentage tariff increase as water is basic need.

There are reasons, however, to assume that the responsemay differ across consumer groups. According to empiricalstudy by Yoo et al. (2014) in Arizona, high-income house-holds tend to respond less than low-income households do.

1We obtain this information from direct correspondence with PT Aetra.

Further, we expect non-household groups, particularly in-dustrial and commercial consumers to be less responsivetoward tariff adjustment as they have less options for PT Ae-tra’s water source substitution. Similarly, we expect that theshort-run response is different from the longer-run response.

To have better understanding of the behavior water con-sumption near the tariff adjustment, Figure 1 illustrates theaverage monthly water consumption from our raw data forhousehold and non-household costumers. The graphs plotaverage monthly of household’s water consumption againstthe month-distance to November 2014. To get better visual-ization, we add a simple regression of the consumption as apolynomial function of time. First, we focus on householdconsumer as it accounts for more than 90% of Aetra’s totalconsumers. There is a positive trend in household’s waterconsumption. Further, the regression lines suggest a drop ofhousehold’s water consumption after November 2014. Sub-stantial drop in water consumption after November 2014 isalso evident for non-household costumers. This result givesan initial evidence of the adverse effect of tariff adjustmenton the water consumption. The subsequent section providesthe empirical evidence of such effect.

4. Results

4.1 Regression estimatesTable presents our panel data estimates of Equation (1). Tocontrol high heterogeneity between group, the result is di-vided into two groups, household and non-household. Firstcolumn of each groups represents the estimate using allsamples. For the evidence on potential heterogeneity, wethen disaggregate two subsamples. Household consumersare divided into R2 and R3–R4 household consumers. Non-household subsamples comprise of commercial-industrialand social consumers. We exclude R1 households and gov-ernment consumers as their number of consumers is rela-tively low.

The result suggests that the price elasticity varies acrosssubcategories. For household consumers, the water con-sumption drops 5.0% on average, statistically different fromzero at 1% level of significance. This result is, however,driven by the R2 household as more than 90% of consumersfalls in this group. Given 14.4% tariff increase for R2 house-hold, the water price elasticity of R2 household is -0.35.If R3 and R4 groups response is estimated separately, thewater consumption drops by 5.6%, yet statistically not dif-ferent from zero. The estimates nevertheless suggest that


http://aat.co.id/informasi-rate/


Figure 2. Average Monthly Water Consumption (m3)Note: red vertical line indicates starting month of new tariff was introduced in November 2014.Blue and red dots represent monthly average water consumption, respectively, measured in m3.

Blue lines are the fitted curves

Table 3. Impacts of Tariff Adjustment on Monthly Water Consumption: Base ResultsHousehold Non-household

All sample R2 R3–R4 All sample Commercial & Industrial Social

Tariff adjustment -0.0510* -0.0505* -0.0578 -0.0401 -0.00824 -0.0755(November 2014) (0.007) (0.007) (0.037) (0.070) (0.088) (0.117)

Month-of-year dummies Yes Yes Yes Yes Yes YesConsumer fixed effects Yes Yes Yes Yes Yes YesQuadratic Polynomial time trend Yes Yes Yes Yes Yes YesAdditional controls Yes Yes Yes Yes Yes Yes# Consumers 52.596 50.243 2.454 2.160 1.808 248# Observations 1,681,917 1,604,295 81,254 49,27 38,609 7,002

Within-R2 0.0613 0.065 0.0187 0.0319 0.0345 0.093Water demand elasticity -0.38* -0.38* -0.43 -0.30 -0.06 -0.56

Note: *, **, *** indicate statistical significance at 1, 5, and 10%, respectively.Clustered Standard errors are reported in parentheses and clustered at consumer-year level.The within-R2 calculates the explanatory power of the time-varying explanatory variables.Price increase is dummy variable, which value is 1 after October 2014, and 0 otherwise.Additional control variables include monthly GDP, monthly precipitation, consumer price index, dummy year, dummy IdulFithri, and dummy drought.

Detail explanation of additional control variables is in Appendix A.

our result is consistent with recent literature on water priceelasticity (for example, see Yoo et al., 2014).

Estimates in Table 3 also suggest that non-householdconsumers are relatively not responsive toward tariff adjust-ment. Tariff adjustment drops the water consumption by3.0%, yet is not different from zero. Similar result is alsofound for the subcategories: social, and commercial andindustrial consumers.

Our table estimates do not provide coefficients for con-trol variables. In this section, we provide a discussion ofthese coefficients. Water consumptions tend to increase byaround 1–3% when precipitation rate increases by one per-cent. We also find GDP elasticity of water consumptionsof 0.3–0.6% for household consumers, statistically differ-ent from zero at 1%. The GDP elasticity is, however, notdifferent from zero for non-household consumers. We findpositive and significant effects of Idul Fithri holiday period,as well as drought in August 2016. Water consumption isseasonal, with the lowest monthly consumptions tending inAugust and the highest in November, ceteris paribus.

Aside from heterogeneity across consumers, there mightbe considerable regional variation in the elasticity. Table 4provides the result for three sub-districts with highest num-ber of customers, i.e. Pasar Kemis, Sepatan, and Cikupa.The estimates are drawn from interacting the interveningvariable and dummy sub-districts. We find substantial varia-

tion of response in the monthly average water consumption.The elasticity of two sub-districts, Pasar Kemis and Sepa-tan, is higher than our baseline estimates of -0.38. Thereare variety reasons that can result in this spatial response.Some sub-districts are relatively easier to find the alterna-tive cheaper water sources, for example ground water, sothat the change in tariff has elastic impact. In contrast, theremight be sub-districts inhabited by substantial number ofnew consumers, for which PT Aetra Air Tangerang offereda flat tariff over certain period, and hence are not affectedby the tariff adjustment.

4.2 Short-run ResponseWe now examine the short-run response of tariff adjustment.The estimates are divided for household and non-householdconsumers. Figure 2 presents the first n-month price elas-ticity. From Panel A, we find that the point estimates ofelasticity are ranging from -0.26 to -0.45, relatively similarwith our household’s estimate in the table 4, and statisticallydifferent from zero at 1% level of significance. Neverthe-less, the magnitude tends to be slightly less elastic for longermonth. Household consumer adjust their water consump-tion more during first several months, yet the effect is lessin the longer-run period.

In Panel B, the short-run response for the non-householdconsumer is within the region of –(0.75–0.00) and all point



Table 4. Heterogeneity Across SubdistrictHousehold

(1) (2) (3)

Tariff adjustment -0.0227* -0.0390* -0.0576*(November 2014) (0.006) (0.006) (0.006)

*Pasar Kemis -0.0572*(0.005)

*Sepatan -0.102*(0.006)

*Cikupa 0.0259*(0.005)

Month-of-year dummies Yes Yes YesConsumer fixed effects Yes Yes YesQuadratic Polynomial time trend Yes Yes YesAdditional controls Yes Yes Yes

# Consumers 52,596# Observations 1,681,917

Within-R2 0.0617 0.0619 0.0614Water demand elasticity -0.53* -0.91* -0.17*

Note: *, **, *** indicate statistical significance at 1, 5, and 10%,respectively. Standard errors are reported in parentheses and clusteredat consumer-year level. The within-R2 calculates the explanatorypower of the time-varying explanatory variables. Price increase isdummy variable, which value is 1 after October 2014, and 0 otherwise.Additional control variables include monthly GDP, monthlyprecipitation, consumer price index, dummy year, dummy IdulFithri, and dummy drought. Detail explanation of additionalcontrol variables is in Appendix A.

estimates are not statistically not different from zero at 5%level of significance. Focusing on the point estimates alone,the magnitude is rather more elastic for longer-run period.

4.3 Testing the RobustnessWe follow Luechinger & Roth (2016) in performing ourrobustness test of our model to changes in time polyno-mial order. The check is done for both household and non-household consumer. Table 5 presents the estimation resultup to fourth time polynomial order. Columns 1–4 are forhousehold estimates, and the rest is for non-household esti-mates.

The estimates are, in general robust in both, sign andmagnitude. Except for the linear order, the estimates arewithin –(0.051–0.057) and –(0.040–0.045) for householdand non-household consumer, respectively. Additionally, allestimates are statistically significant at 1% level for house-hold consumer, consistent with the base estimates in mainresult. Similar result is also found for non-household con-sumer: all estimates are statistically insignificant.

Our robustness check provides, however, less intuitiveresult for higher polynomial order. Both magnitude and sta-tistical significance substantially change as the order getshigher, as presented in the Appendix. This indicates that es-timates are substantially affected by the higher polynomialorder (Luechinger & Roth, 2016). We guess that our timeframe, which is relatively much shorter than other similarregression-discontinuity studies is partially responsible forinconsistent estimates for higher order.

To further check the robustness of the estimates, we usefalsification tests by shifting the tariff adjustment one andto year backward to October 2013 and 2012, respectively.Table 6 reports the results for household and non-householdconsumers. The estimates become positive and statistically

significant for household consumers, which are less intuitive.Meanwhile, the estimates for non-household consumers arenegative and statistically not different from zero.

5. Discussion

In this paper, we have estimated water demand elasticitiesusing unprecedented tariff adjustment event in Tangerangregency, Indonesia. We used regression discontinuity frame-work using time as forcing variable to elicit the casual im-pact between the tariff change and monthly water consump-tion. This approach avoids us from the endogeneity prob-lem or estimation complexity if average or marginal priceis used, respectively. The estimates are relatively robust.The robustness check by time polynomial also shows theestimated elasticities are mostly consistent to be less thanunity or statistically not different from zero.

Our estimates results show that monthly water demandelasticities of PT Aetra Air Tangerang household consumersin Tangerang regency of Indonesia is -0.38 on average.Higher-income tends to be not sensitive toward tariff ad-justment policy. This result confirms recent internationalevidence that the elasticities are likely to be less than zero.Specifically, the results are within overview of the estimatesfor developing countries by Nauges & Whittington (2009),which lie in the range from -0.3 to -0.6. Our estimates arealso close with Dalhuisen et al.’s (2003) meta-analysis ofprice elasticities of water demand between 80’s and 90’s.For similar estimates in Indonesia, our estimates are muchlower than Rietveld et al.’s (2000) elastic estimates for Solo,Indonesia. Table 7 summarizes selected prior estimates ofwater demand elasticity. For household consumers, our es-timates are relatively comparable with other similar stud-ies. The estimates are slightly higher than Nataraj & Hane-



Figure 3. Monthly Short-run Response

mann’s (2011) estimates of -0.12 for Santa Cruz, California,Mansur & Olmstead’s (2012) of -0.33 for 11 North Ameri-can cities, Polycarpou & Zachariadis’ (2013) of -0.25 forCyprus, and Strong & Goemans’ (2015) estimates of -0.23for Aurora, Colorado. Our estimates are relatively compara-ble with Martınez-Espineira’s (2007) for Spain. Yoo et al.(2014) reported much elastic result for the case of Phoenix,Arizona. In addition, our short-run estimates are slightlyhigher than the long-run estimates, despite remaining inelas-tic. Martınez-Espineira (2007) used error-correction modelto distinguish short-run and long-run water demand elas-ticity for Spain and found that his short-run estimates aremore inelastic than the long-run estimates.

We also found that non-household consumers are notsensitive toward the tariff adjustment. Meanwhile, we findno statistical difference of response across time for non-household consumers. Yet, our short-run estimates for non-household consumers are relatively lower than Shen & Lin’s(2017) estimates of -0.12 for agricultural water demand inChina or Davidson & Hellegers’ (2011) estimates of -0.44for irrigation water in Musi, India.

Our estimates of water demand elasticity could be usedto parameterize the aggregate demand for water and thusto project the water demand-supply balance in the future.ADB (2016) reported that current local government ownwater utilities only cover 30–40% of the service area and thetrend in gap is widening. Given the estimated elasticities,the government may formulate pricing strategies for waterutilities to manage the demand, while at the same timeexpand the service.

References

ADB. (2016). Indonesia Country Water Assessment. Manila:Asian Development Bank. https://www.adb.org/sites/default/files/institutional-document/183339/ino-water-assessment.pdf

Auffhammer, M., & Kellogg, R. (2011). Clearing the air?The effects of gasoline content regulation on air qual-ity. American Economic Review, 101(6), 2687–2722. doi:10.1257/aer.101.6.2687.

Badan Pusat Statistik. (2017a). Gross domestic product 2010-2013,Indonesia [[Seri 2010] PDB Triwulanan Atas Dasar HargaKonstan 2010 Menurut Lapangan Usaha (Miliar Rupiah), 2010-2013]. https://www.bps.go.id/linkTableDinamis/view/id/1135.

Badan Pusat Statistik. (2017b). Gross domestic product 2014-2017,Indonesia [[Seri 2010] PDB Triwulanan Atas Dasar HargaKonstan 2010 Menurut Lapangan Usaha (Miliar Rupiah), 2014-2017]. https://www.bps.go.id/dynamictable/2015/05/06/827/-seri-2010-pdb-triwulanan-atas-dasar-harga-konstan-2010-menurut-lapangan-usaha-miliar-rupiah-2014-2017.html.

Badan Pusat Statistik. (2017c). Monthly consumer price in-dex, Indonesia [Jumlah Desa/Kelurahan Menurut Provinsidan Topografi Wilayah, 2003 - 2014]. https://www.bps.go.id/linkTableDinamis/view/id/907.

Bertrand, M., Duflo, E., & Mullainathan, S. (2004). Howmuch should we trust differences-in-differences estimates?.The Quarterly Journal of Economics, 119(1), 249–275. doi:https://doi.org/10.1162/003355304772839588.

Burke, P. J., Shahiduzzaman, M., & Stern, D. I. (2015). Carbondioxide emissions in the short run: The rate and sources ofeconomic growth matter. Global Environmental Change, 33,109–121. doi: https://doi.org/10.1016/j.gloenvcha.2015.04.012.

Burke, P. J., Batsuuri, T., & Yudhistira, M. H. (2017). Easing thetraffic: The effects of Indonesia’s fuel subsidy reforms on toll-


https://www.adb.org/sites/default/files/institutional-document/183339/ino-water-assessment.pdf

https://www.adb.org/sites/default/files/institutional-document/183339/ino-water-assessment.pdf

https://www.bps.go.id/linkTableDinamis/view/id/1135

https://www.bps.go.id/dynamictable/2015/05/06/827/-seri-2010-pdb-triwulanan-atas-dasar-harga-konstan-2010-

https://www.bps.go.id/dynamictable/2015/05/06/827/-seri-2010-pdb-triwulanan-atas-dasar-harga-konstan-2010-

menurut-lapangan-usaha-miliar-rupiah-2014-2017.html




Table 5. Robustness Check: Results Under Different Polynomial OrderHousehold Non-Household

1st 2nd 3rd 4th 1st 2nd 3rd 4th

Tariff adjustment -0.103* -0.051* -0.057* -0.055* -0.195* -0.040 -0.041 -0.045(November 2014) (0.006) (0.007) (0.007) (0.007) (0.065) (0.070) (0.071) (0.071)

Month-of-year dummies Yes Yes Yes Yes Yes Yes Yes YesConsumer fixed effects Yes Yes Yes Yes Yes Yes Yes YesQuartic Polynomial time trend Yes Yes Yes Yes Yes Yes Yes YesAdditional controls Yes Yes Yes Yes Yes Yes Yes Yes

# Consumers 52.596 2.160# Observations 1.681,92 49.270

Within-R2 0.061 0.061 0.061 0.061 0.030 0.032 0.032 0.032Water demand elasticity -0.75* -0.38* -0.42* -0.41* -1.36* -0.30 -0.31 -0.34

Note: *, **, *** indicate statistical significance at 1, 5, and 10%, respectively.Standard errors are reported in parentheses and clustered at consumer-year level.The within-R2 calculates the explanatory power of the time-varying explanatory variables.Price increase is dummy variable, which value is 1 after October 2014, and 0 otherwise.Additional control variables include monthly GDP, monthly precipitation, consumer price index, dummy year, dummyIdul Fithri, and dummy drought.


Table 6. Falsification Tests: Shifting the Date of Tariff InterventionHousehold Non-Household

October 2013 October 2012 October 2013 October 2012

Tariff increase 0.00927** 0.0243* -0.0514 -0.145(0.005) (0.008) (0.058) (0.119)

Month-of-year dummies Yes Yes Yes YesConsumer fixed effects Yes Yes Yes YesQuadratic Polynomial time trend Yes Yes Yes YesAdditional controls Yes Yes Yes Yes# Consumers 52,596 52,596 2,16 2,16# Observations 1,681,917 1,681,917 49,27 49,27

Within-R2 0.00869 0.0087 0.026 0.026Note: *, **, *** indicate statistical significance at 1, 5, and 10%, respectively.

Standard errors are reported in parentheses and clustered at consumer-year level.The within-R2 calculates the explanatory power of the time-varying explanatory variables.Price increase is dummy variable, which value is 1 after October 2014, and 0 otherwise.Additional control variables include monthly GDP, monthly precipitation, consumer price index,dummy year, dummy Idul Fithri, and dummy drought.


road travel. Transportation Research Part A: Policy and Practice,105, 167–180. doi: https://doi.org/10.1016/j.tra.2017.08.003.

Cameron, A. C., Gelbach, J. B., & Miller, D. L. (2011).Robust inference with multiway clustering. Journal ofBusiness & Economic Statistics, 29(2), 238–249. doi:https://doi.org/10.1198/jbes.2010.07136.

CEIC. (2017). CEIC data. www.ceicdata.com.Chen, Y., & Whalley, A. (2012). Green infrastructure: The effects

of urban rail transit on air quality. American Economic Journal:Economic Policy, 4(1), 58–97. doi: 10.1257/pol.4.1.58.

Dalhuisen, J. M., Florax, R. J., De Groot, H. L., & Nijkamp,P. (2003). Price and income elasticities of residential waterdemand: a meta-analysis. Land Economics, 79(2), 292–308.doi: 10.2307/3146872.

Davidson, B., & Hellegers, P. (2011). Estimating the own-priceelasticity of demand for irrigation water in the Musi catch-ment of India. Journal of Hydrology, 408(3–4), 226–234. doi:https://doi.org/10.1016/j.jhydrol.2011.07.044.

Geospatial Information Agency. (2014). Map of Tangerang Re-gency. www.big.go.id.

Hewitt, J. A., & Hanemann, W. M. (1995). A discrete/continuouschoice approach to residential water demand underblock rate pricing. Land Economics, 71(2), 173–192. doi:10.2307/3146499.

Jales, H., & Yu, Z. (2017). Identification and Estimation Using

a Density Discontinuity Approach. In M. D. Cattaneo & J. C.Escanciano (eds.), Regression Discontinuity Designs (Advancesin Econometrics, Volume 38), pp. 29–72, Emerald PublishingLimited.

Keputusan Bupati Tangerang No. 690/Kep.206.Huk/2017 (March27, 2017).

Lee, D. S., & Lemieux, T. (2010). Regression discontinuity designsin economics. Journal of Economic Literature, 48(2), 281–355.doi: 10.1257/jel.48.2.281.

Luechinger, S., & Roth, F. (2016). Effects of a mileage taxfor trucks. Journal of Urban Economics, 92, 1–15. doi:https://doi.org/10.1016/j.jue.2015.09.005.

Lundqvist, J., Appasamy, P., & Nelliyat, P. (2003). Dimensionsand approaches for Third World city water security. Philosoph-ical Transactions of the Royal Society B: Biological Sciences,358(1440), 1985–1996. doi: 10.1098/rstb.2003.1382.

Mansur, E. T., & Olmstead, S. M. (2012). The value ofscarce water: Measuring the inefficiency of municipal regu-lations. Journal of Urban Economics, 71(3), 332–346. doi:https://doi.org/10.1016/j.jue.2011.11.003.

Martinez-Espineira, R. (2002). Residential water de-mand in the Northwest of Spain. Environmentaland Resource Economics, 21(2), 161–187. doi:https://doi.org/10.1023/A:1014547616408.

Martınez-Espineira, R. (2007). An estimation of residential water


www.ceicdata.com

www.big.go.id


Table 7. Summary of Selected Prior Estimates of Water Demand Elasticity

Study Country Price elasticity of water demand NoteShort-run Long-run

Hewitt and Hanemann (1995) US -1.60 ResidentialRietveld et al. (2000) Indonesia -1.20 ResidentialMartinez-Espineira (2002) Spain -0.12 ResidentialMartinez-Espineira (2007) Spain -0.10 -0.50 ResidentialDavidson and Hellegers (2011) India -0.44 IrrigationNataraj and Hanemann (2011) US -0.12 ResidentialStrong and Goemans (2015) US -0.23 ResidentialPolycarpou and Zachariadis (2013) Cyprus -0.25 ResidentialYoo et al. (2014) US -0.90 ResidentialShen and Lin (2017) China -0.12 Agricultural

Note: Studies are ordered chronologically.The list is sample studies only.Definition of short-run and long-run may differ across studies.

demand using co-integration and error correction techniques.Journal of Applied Economics, 10(1), 161–184.

NASA. (2017). Giovanni v 4.20.5. National Aeronautics and SpaceAdministration. http://giovanni.sci.gsfc.nasa.gov/giovanni.

Nataraj, S., & Hanemann, W. M. (2011). Does marginalprice matter? A regression discontinuity approachto estimating water demand. Journal of Environmen-tal Economics and Management, 61(2), 198-212. doi:https://doi.org/10.1016/j.jeem.2010.06.003.

Nauges, C., & Whittington, D. (2009). Estimation of wa-ter demand in developing countries: An overview. TheWorld Bank Research Observer, 25(2), 263–294. doi:https://doi.org/10.1093/wbro/lkp016.

Patunru, A. A. (2015). Access to safe drinking water and sanitationin Indonesia. Asia & the Pacific Policy Studies, 2(2), 234–244.doi: 10.1002/app5.81.

Polycarpou, A., & Zachariadis, T. (2013). An econometric analysisof residential water demand in Cyprus. Water Resources Man-agement, 27(1), 309–317. doi: https://doi.org/10.1007/s11269-012-0187-x.

Rietveld, P., Rouwendal, J., & Zwart, B. (2000). Block ratepricing of water in Indonesia: an analysis of welfare effects.Bulletin of Indonesian Economic Studies, 36(3), 73–92. doi:10.1080/00074910012331338983.

Shen, X., & Lin, B. (2017). The shadow prices and demand elas-ticities of agricultural water in China: A StoNED-based analy-sis. Resources, Conservation and Recycling, 127, 21–28. doi:https://doi.org/10.1016/j.resconrec.2017.08.010.

Strong, A., & Goemans, C. (2015). The impact of real-time quantity information on residential water de-mand. Water Resources and Economics, 10, 1–13. doi:https://doi.org/10.1016/j.wre.2015.02.002.

York, R. (2012). Asymmetric effects of economic growth anddecline on CO2 emissions. Nature Climate Change, 2(11), 762–764. doi: 10.1038/nclimate1699.

Yoo, J., Simonit, S., Kinzig, A. P., & Perrings, C. (2014). Estimat-ing the price elasticity of residential water demand: the case ofPhoenix, Arizona. Applied Economic Perspectives and Policy,36(2), 333–350.


http://giovanni.sci.gsfc.nasa.gov/giovanni


Appendix A: Variable Definitions

Water consumption: Number of monthly water consumption in m3. Source: direct correspondence with PT Aetra AirTangerang.

GDP: monthly gross domestic product. The data is available at quarterly and yearly basis. We follow Burke et al. (2017)approach to construct the variable by weighting the yearly GDP by that month’s industrial production index relativeto other months in respective year. The formula is:

(A1)GDPm = GDPy ∗IndustrialProductionIndexm

∑12ndmontho f year1st montho f year IndustrialProductionIndexi

Source: Badan Pusat Statistik (2017a,b) and CEIC (2017)Precipitation: area-averaged monthly precipitation at village level, measured in millimeters. Source: National Aeronautics

and Space Administration/NASA (2017), series TRMM 3B43.CPI: monthly national average consumer price index. Source: Badan Pusat Statistik (2017c)Idul Fithri: dummy variable that takes 1 if a respective month contains Idul Fithri holidays, and 0 otherwise. Source:

www.timeanddate.comFollowing Idul Fithri: dummy variable that takes 1 for months after the Idul Fithri holidays, and zero otherwise. Source:

Source: www.timeanddate.comDrought: dummy variable representing drought event in Cisadane river, the main supplying raw water for PT Aetra Air

Tangerang, at August 2015. The variable takes 1 for August 2015, and 0 otherwise. Source: direct correspondencewith PT Aetra Air Tangerang.


lpem-feb ui working paper 018, march 2018 issn 2356-4008 · are served by pdam of tangerang...

Documents